Protocols

The Protocols section of the BioLM Console provides end-to-end workflows that combine multiple BioLM models to perform more complex experimental objectives.

Each protocol represents a defined, reproducible procedure—similar to a wet-lab protocol—but executed through BioLM’s AI-driven molecular modeling stack.

Protocols automate tasks such as variant generation, folding, scoring, and validation across integrated model pipelines.

Overview

Each protocol card displays:

Protocol Name – the workflow identifier.
Description – a concise summary of the scientific objective and models involved.
Run Protocol – launches the workflow with default or user-provided parameters.
You can also create custom protocols by clicking Create New Protocol or view past runs under View Activity.

Example: AntiFold Design Antibodies

The AntiFold Design Antibodies protocol provides an end-to-end pipeline for antibody variant generation and evaluation.

It integrates several models—including AntiFold, ABodyBuilder, and IgBert—to design, fold, and score antibody sequences in one automated workflow.

Steps to Run

From the Protocols page, select AntiFold Design Antibodies and click Run Protocol.
Under Configure Your Protocol, provide required parameters:
- Heavy Chain ID and Light Chain ID (e.g., H , L )
- Antigen Chain ID (if applicable)
- Number of Samples to generate
- Temperature (controls sampling diversity)
- Regions to target (e.g., CDR1–3)
- Optionally enable SADIE Validation for structure verification
Upload or paste a PDB file for the antibody-antigen complex.
Click Run Protocol to start the workflow. You can load example data for demonstration purposes.

Monitoring a Protocol Run

Once launched, you’ll be redirected to the Protocol Activity View, which displays live progress and results.

Pipeline Stages

Each stage represents a step in the workflow:

Generation – produces new sequence variants using AntiFold inverse-folding.
Sequence Batch Scoring – evaluates generated variants with integrated scoring models.
Folding – predicts 3D structures of top candidates.
Scoring – assigns final pLDDT and log-probability metrics.
Analysis – aggregates validation data and identifies best-performing variants.
Gather Designed PDBs – collects final structural outputs.
Compute RMSD – calculates deviation between predicted and reference structures.

The Tasks tab lists sub-tasks for each stage, showing live completion, errors, and performance statistics.

Tabs Overview

Output – Displays top-scoring variants, sortable by metrics such as RMSD, pLDDT, and global score. Results can be viewed as tables or cards, and exported as CSV or JSON.
Tasks – Shows workflow progress and completion status for each task (running, completed, failed).
Inputs – Lists input parameters used for the run (e.g., regions, number of samples, temperature).
Logs – Displays real-time runtime logs for debugging and monitoring.

Reviewing Results

Completed runs display Variant Results, with ranked sequences and associated metrics:

Score – combined global quality score
RMSD / pLDDT – structure confidence metrics
Mutations – number of residue changes from the template
Log Probabilities – sequence likelihood estimates
PDB Structures – downloadable designed models

You can toggle between Table and Card views, or click Export Data to save the outputs.

Viewing Activity

Under Activity → Protocols, you can:

Review recent and historical protocol runs
Check run status (Running, Succeeded, Failed)
Re-open or duplicate previous runs for comparison or re-execution

Each run entry links back to its detailed results page and includes a unique Run ID.

Notes

Protocols may use multiple BioLM models concurrently and require available compute credits.
Workflows are containerized for reproducibility and can be re-run with modified inputs.
Results, logs, and derived models are stored securely within your workspace environment.
Some advanced protocols may require access to team or premium plans for full functionality.